Pharmacokinetics-Pharmacodynamics Analysis of Bicyclic 4-Nitroimidazole Analogs in a Murine Model of Tuberculosis

PA-824 is a bicyclic 4-nitroimidazole, currently in phase II clinical trials for the treatment of tuberculosis. Dose fractionation pharmacokinetic-pharmacodynamic studies in mice indicated that the driver of PA-824 in vivo efficacy is the time during which the free drug concentrations in plasma are above the MIC (fT>MIC). In this study, a panel of closely related potent bicyclic 4-nitroimidazoles was profiled in both in vivo PK and efficacy studies. In an established murine TB model, the efficacy of diverse nitroimidazole analogs ranged between 0.5 and 2.3 log CFU reduction compared to untreated controls. Further, a retrospective analysis was performed for a set of seven nitroimidazole analogs to identify the PK parameters that correlate with in vivo efficacy. Our findings show that the in vivo efficacy of bicyclic 4-nitroimidazoles correlated better with lung PK than with plasma PK. Further, nitroimidazole analogs with moderate-to-high volume of distribution and Lung to plasma ratios of >2 showed good efficacy. Among all the PK-PD indices, total lung T>MIC correlated the best with in vivo efficacy (rs = 0.88) followed by lung Cmax/MIC and AUC/MIC. Thus, lung drug distribution studies could potentially be exploited to guide the selection of compounds for efficacy studies, thereby accelerating the drug discovery efforts in finding new nitroimidazole analogs.


Introduction
Every year nearly 8 million new cases of Tuberculosis (TB) are reported globally resulting in 1.4 million deaths [1]. Poor treatment compliance -due to the requirement for prolonged multidrug therapy -as well as the use of inadequate regimens has fueled the emergence of multi-drug-resistant and extensively-drugresistant (MTD-TB and XDR-TB) TB strains. MDR-TB is resistant to at least isoniazid and rifampicin and XDR-TB is resistant to isoniazid, rifampicin, fluoroquinolones and at least one of the injectables [2]. TB control programs are further complicat-ed in settings where the incidence of co-infection with HIV is high because drug-drug interactions with anti-retroviral therapy are difficult to avoid [2,3]. Hence there is an urgent need to discover new TB drugs active against drug-resistant forms of TB and compatible with treatment against HIV.
PA-824 [4] and OPC-67683 [5] are two bicyclic 4-nitroimidazoles currently in phase II clinical trials, representing a promising new class of therapeutics for TB [6]. Preclinical testing of PA-824 showed bactericidal activity in various in vitro and in vivo models [7,8]. PA-824 was shown to be well tolerated in healthy subjects, following oral daily doses for 7 days [9]. These results, combined with the demonstrated activity of PA-824 against drug-sensitive and multidrug-resistant Mtb, supported the progression of this compound and its evaluation as a novel treatment against TB. An early bactericidal activity (EBA) study of PA-824 revealed a lack of dose response between 200 and 1200 mg administered daily for 14 days [10]. Dose-fractionation PK-PD studies in mice showed the PK-PD driver of PA-824 to be the time during which the free drug concentrations in plasma were above the MIC (fT .MIC ) [11]. In retrospect, clinical investigators established that fT .MIC was 100% at all doses between 200 and 1200 mg daily. An additional phase II trial between 50 and 200 mg was undertaken to establish the lowest efficacious dose [12]. 200 mg of PA-824 was found to be efficacious and used in combination with other anti-TB drugs [13].
Physico-chemical properties, in vitro potency, in vitro and in vivo pharmacokinetics (PK) are critical determinants for in vivo efficacy. PA-824 is highly lipophilic and exhibits poor aqueous solubility. To overcome the limitation of its low solubility and improve its oral bioavailability, a cyclodextrin formulation was developed and used for in vivo animal efficacy studies [4,7]. Extensive lead optimization efforts were undertaken to improve aqueous solubility, metabolic stability, in vitro potency and in vivo efficacy of anti-tubercular nitroimidazoles and various analogs were synthesized [5,[14][15][16][17][18][19][20][21][22][23]. Comprehensive in vivo pharmacology studies are generally resource and time intensive. This is particularly true for TB because of the slow rate of M. tuberculosis growth, lengthy treatment duration and requirement of high containment facility. In this study, a panel of closely related potent bicyclic 4-nitroimidazoles (NI) was profiled both in vitro and in vivo. The data is retrospectively analyzed to identify the PK parameters that correlate with in vivo efficacy of a series of bicyclic 4-nitroimidazoles. The results of this investigation showed that PK properties such as volume of distribution and lung exposure predicts in vivo efficacy of bicyclic 4-nitroimidazoes better than other PK parameters. Thus, in vitro potency and lung PK could be used as surrogate to guide the prioritization of new pre-clinical candidates for lengthy efficacy studies, there by expediting drug discovery.

In vitro physico-chemical properties
In vitro physicochemical and PK parameters like solubility, log P, PAMPA, Caco-2 permeability and mouse plasma protein binding were determined in-house in medium to high throughput format assays. Briefly, solubility was measured using a high throughput equilibrium-solubility (HT-Eq sol) assay using a novel miniaturized shake-flask approach and streamlined HPLC analysis [25]; lipophilicity determination was carried out in 96-well micro titer plates and the diffusion of compounds between two aqueous compartments separated by a thin octanol liquid layer was measured [26]; PAMPA permeability experiments were carried out in 96-well micro titer filter plates at absorption wavelengths between 260 and 290 nm [27]; Caco-2 assay was carried out in a 96-well format, and compound concentrations in each chamber were measured by LC/MS as described previously [28] and plasma protein binding was determined in mouse plasma using an ultra-filtration method [29].

Ethics Statement
All animal experimental protocols (protocol #023/2009 and #025/2009 for PK; protocol #004/2010 for efficacy) involving mice were reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) of Novartis Institute for Tropical Diseases (NITD). The animal research complied with Singapore Animal Veterinary Authority and global Novartis policy on the care and use of animals. Experimental and control animals infected with Mtb were euthanized at the end of the experiment. All procedures during pharmacokinetic experiments were performed under isoflurane anesthesia and all efforts were made to minimize suffering.

In vivo pharmacokinetic (PK) studies
Female CD-1 mice obtained from Biological Resource Center in Singapore were used for in vivo PK studies. Mice were acclimatized before initiation of pharmacokinetic (PK) experiments. Feed and water were given ad libitum. The compounds were formulated at a concentration of 1, 2.5 or 5 mg/mL for a dose of 10 mg/kg (Amino-824, AminoEthyl-824) or 25 mg/kg (PA-824, NI-135, NI-147, NI-136, NI-176, NI-269, NI-182, NI-145, NI-297 and NI-302) or 50 mg/kg (NI-622 and NI-644) given orally and at 1 or 2 mg/mL concentration for a dose of 5 or 10 mg/kg given intravenously. The CM-2 formulations were prepared in 10% w/v hydroxypropyl-b-cyclodextrin and 10% v/v lecithin in water as described earlier [7,8]. The formulation was centrifuged and the supernatant was collected for intravenous administration. After oral dosing, blood and lung samples from mice were collected at various time points ranging from 0.08 hrs (but 0.02 hrs for i.v dosing) to 48 hrs. Groups of three mice were used for each time point. Blood was centrifuged at 13,000 rpm for 7 min at 4uC, plasma was harvested and stored at -20uC until analysis. Lung tissue samples were excised, dipped in PBS, gently blotted with absorbent paper, dried, weighed and stored at -20uC until further analysis.
For LC/MS/MS analysis, 50 mL of plasma samples were precipitated using 400 mL of acetonitrile:methanol:acetic acid (90:9.8:0.2) containing 500 ng/mL of either related compound or warfarin as internal standard. After vortexing and centrifuging the mixture, the supernatant was removed and 5 mL of sample analyzed. Whole lung tissue was homogenized in 2 mL of PBS. 50 mL of lung homogenate was taken and processed as described above for plasma samples. The standard calibration curve was prepared by spiking blank plasma and lung tissue with different concentrations of the compound. In addition, quality control samples with three different concentrations were prepared in respective blank matrix and analyzed together with the unknown samples for validation purposes. Analyte quantitation was performed by LC/MS/MS using optimized conditions for each compound. Liquid chromatography was performed using an Agilent 1100 HPLC system (Santa Clara, CA), with the Agilent Zorbax XDB Phenyl (3.5 m, 4.6675 mm) column at an oven temperature of 35uC and 45uC, coupled with a triple quadruple mass spectrometer (Applied Biosystems, Foster City, CA). Instrument control and data acquisition were performed using Applied Biosystems software Analyst 1.4.2. The mobile phases used were A: water-acetic acid (99.8:0.2, v/v) and B either as: acetonitrileacetic acid (99.8:0.2, v/v) or methanol-acetic acid (99.8:0.2, v/v), using a gradient, with a flow rate of 1.0 mL/min, and a run time of 6 to 8 min. Under these conditions the retention times of various compounds ranged between 3.2 and 6.5 minutes. Multiple reaction monitoring (MRM) was combined with optimized mass spectrometry parameters to maximize detection specificity and sensitivity. Most of the compounds were analyzed using electrospray ionization in the positive mode. The recovery of the compound from both plasma and lung tissue were good and consistent across the concentration range studied. The lower limit of quantification for different compounds ranged between 1 and 49 ng/mL in plasma and 1 and 132 ng/g in lungs. Calibration curve was prepared freshly and analyzed with every set of study samples. Intraday variability was established with triplicate quality control samples at three concentration levels. The results were accepted if relative standard deviation was less than 15%.
Mean values of compound concentrations in plasma and lungs were obtained from three animals at each time point and plotted against time to generate concentration-time profiles. PK parameters were determined using Phoenix WinNonlin, version 6.3 (Pharsight, A Certara company, USA, www.pharsight.com), by non-compartmental modeling using built-in model (200-202) for extravascular and intravenous bolus dosing. The oral bioavailability (F) was calculated as the ratio between the area under the curve (AUC inf ) following oral administration and the AUC inf following intravenous administration corrected for dose (

In vivo mouse efficacy studies
In vivo mouse efficacy studies were determined after intranasal infection of Balb/c mice with 10 3 cfu Mtb H37Rv. Treatment was initiated 4 weeks after infection. Compounds were orally administered in CM-2 formulation for 4 weeks daily. Bacterial loads were determined at 2 and 4 weeks post treatment [30]. Statistical analysis was done by a one-way analysis of variance, followed by a multiple comparison analysis of variance by a oneway Tukey test (GraphPad Prism, version 5.02, San Diago, California USA, www.graphpad.com). Differences were considered statistically significant at the 95% level of confidence [7].

Calculation of PK-PD parameters
The MIC against Mtb was used to calculate PK-PD indices (C max /MIC, AUC/MIC and T .MIC ). The C max /MIC was defined as the ratio of peak plasma concentration (C max ) to the MIC, the AUC/MIC was defined as the ratio of the AUC 0-24 to the MIC, and the time above MIC (T .MIC ) was defined as 24 h period during which the total compound concentration exceeded the MIC. C max /MIC and AUC/MIC were calculated as ratios from PK parameter obtained from non-compartmental analysis and MIC. T .MIC were derived from Phoenix WinNonlin software by specifying MIC as therapeutic response and time above therapeutic response was obtained. Using plasma protein binding, unbound concentrations in plasma were calculated, PK parameters were derived from Phoenix WinNonlin and PK-PD indices were defined as fC max /MIC, fAUC/MIC and % fT .MIC where 'f' refers to free concentration. For calculation and plotting of mean concentration-time curve, concentrations indicated as below the lower limit of quantification (LLOQ) were replaced by 0.5*LLOQ. Ignoring the values here would impact some of the PK-PD parameters. This approach has no impact on pharmacokinetic parameter calculations [31].

PK-PD analysis
PK-PD indices were estimated from the plasma and lung drug concentrations, in vitro potency and plasma protein binding. A Spearman's rank correlation [32,33] was run to determine the relationship between various PK parameters and mean log CFU reduction using Prism software (GraphPad Prism, version 5.02, San Diago, California USA, www.graphpad.com).

In vitro potency and physico-chemical properties
In an effort to improve the solubility and potency of PA-824, diverse structural analogs of PA-824 were synthesized and their in vitro activities reported [14,16,[18][19][20]. A few potent bicyclic 4nitroimidazole analogs were selected and characterized in detail ( Figure 1). In vitro Mtb potency and physico-chemical properties of these nitroimidazole analogs are summarized in Table 1. All the NI analogs studied showed Mtb specific growth inhibitory activity and no cytotoxicity was observed in THP1 macrophage cell lines ( Table 1). F 420 deficient (FbiC) mutants were resistant to all these bicyclic 4-nitroimidazoles analogs ( Table 1), suggesting a mechanism of action similar to PA-824, involving F 420 -dependent bioactivation [24]. Modifications on the benzyl ring (NI-135, NI-147, NI-136 and NI-176), and oxazine ring (NI-269, NI-182, NI-145, NI-297, NI-302 and NI-176) showed significant improvement of in vitro potency compared to PA-824. The nitroimidazole (NI) analogs tested in this study displayed a wide range of solubility (,2 to 127 mg/mL). Amino-nitroimidazoles showed improved solubility compared to their respective benzyl ether analogs (Amino-824 vs. PA-824 and NI-269 vs. NI-145). NI-297, a biphenyl derivative of NI-182, had very poor aqueous solubility (,2 mg/mL) due to its high lipophilicity (logP 6). In general, the logP of all the other NI derivatives ranged between 2.4 and 3.8 and all showed high permeability except NI-644, which had moderate permeability in the Caco-2 assay. Overall the compounds exhibited moderate-tohigh mouse plasma protein binding (80 to 98%), except for Aminoethyl-824, which showed the lowest binding (45%).

In vivo plasma PK properties
Each compound was subjected to intravenous and oral mouse PK in CM-2 formulation. The total compound concentration in mouse plasma was measured and free plasma concentrations were calculated using in vitro plasma protein binding (Table 1 and 2). NI-147 with a hydroxyl functional group on the benzyl ring, displayed markedly inferior PK properties (Table 2, Figure S1 in File S1). The poor PK is likely due to glucuronidation of the hydroxyl group as suggested by the presence of an extra peak corresponding to +176 Da in the mass spectrometry analysis. Hence NI-147 was not included in in vivo mouse efficacy studies. The NI analogs displayed a wide range of volume of distribution (Vss = 0.7 to 4.2 L/kg) corresponding to 1.1 to 7 times total body water. NI-135 showed higher Vss (4.2 L/kg), followed by NI-182, NI-297 and NI-269 (2.6 to 3 L/kg). All other analogs showed moderate Vss similar to PA-824, except for NI-644, which showed a low volume of distribution (Vss = 0.7 L/kg). The total systemic clearance was low to moderate (4 to 44 mL/min/kg) corresponding to 5 to 49% of hepatic blood flow. NI-135 and AminoEhtyl-824 showed moderate clearance (41 and 44 mL/min/kg respectively). All other analogs showed clearance similar to PA-824 (10 to 25 mL/min/kg), except for NI-297 and NI-302, which showed very low clearance (5 mL/min/kg). The elimination half-life ranged between 0.7 h and 6.7 h for the NI analogs studied. NI-297, NI-135 and NI-302 showed long half-life (3.7 to 6.7 h). All other analogs showed half-life similar to PA-824 (1.3 to 2.8 h), except for NI-644 and AminoEthyl-824, which showed short halflife (,1 h). Generally, the NI analogs at comparable doses displayed rapid absorption (T max , 0.3 to 4 h), except for NI-297, which showed delayed absorption with a T max of 8 h, possibly due to its higher lipophilicity and lower solubility. The peak plasma concentration (C max ) ranged between 1.2 mg/mL and 12.9 mg/mL and exposure (AUC) ranged between 4.8 mg.h/mL and 144 mg.h/ mL (Table 2). NI-135 had significantly lower plasma C max , exposure and oral bioavailability compared to PA-824, likely due to its three times higher in vivo clearance combined with higher volume of distribution. On the contrary, NI-302 and NI-297 had higher systemic plasma exposure mostly due to decreased in vivo clearance (Table 2, Figure S1 in File S1). At comparable dose, NI-622 and NI-644 showed similar plasma C max , exposure and oral bioavailability compared to PA-824. All other NI analogs showed moderate plasma exposure and oral bioavailability (64 to 88%) except for NI-135, NI-297 and NI-302. Despite, NI analogs displayed varied aqueous solubility (,2 to 127 ug/mL), in CM2 formulation all the analogs showed moderate to high oral bioavailability (Table 2). Interestingly, NI-145 and NI-297 had least solubility (,2 mg/mL), in CM-2 formulation both compounds showed good oral bioavailability. The use of CM-2 (lipid coated cyclodextrin complexation) formulation is known to improve solubility and bioavailability [34]. At 25 mg/kg, the free plasma C max and AUC parameters ranged from 0.1-0.8 mg/mL and 0.4-5.1 mg.h/mL respectively. NI-644 showed similar free plasma concentration as PA-824, whereas all other NI analogs showed relatively lower free plasma C max and exposure (Table S1 in File S1).

In vivo lung PK properties
The primary and the most important site of Mtb infection in patients is lung tissue. To understand the effect of structural changes in the NI molecules on lung PK parameters, we measured total compound concentration in mouse lungs ( Table 3). The NI analogs at 25 mg/kg dose showed a wide range of values for lung C max (4.2-17.8 mg/g), T max (0.08-2 h) and exposure (18.6-233 mg*h/g). All NI analogs displayed near parallel concentration-time profile in plasma and lung tissue, suggesting a rapid equilibrium between these two tissues. Interestingly, the lung-toplasma partitioning varied from 0.5 to 4.6 for C max and 0.4 to 3.9 for AUCs across the series in correlation with the observed volume of distribution (Table 3). NI-135, NI-136, NI-182 and NI-297 showed lung partitioning of 2.5 to 4.6 fold, and are comparable to PA-824. NI-622, NI-644, Amino-824 and NI-302 showed lower lung partitioning (,1) compared to PA-824. Although, NI-135 and NI-136 showed higher lung to plasma ratio (3.6 to 4.6), their absolute lung concentrations were 2.5 to 7.5 fold lower than PA-824. On the contrary, NI-302 showed lower lung partitioning, but its absolute concentrations were comparable to PA-824 (Table 3).

Established mouse efficacy
Based on the in vitro potency and the in vitro and in vivo PK results, ten compounds were selected for in vivo mouse efficacy studies with 4 weeks of daily oral treatment. The mean lung CFU reductions compared to untreated controls are summarized in Table 4. The efficacy ranged from 0.5 to 1.56 log at 25 mg/kg and 0.6 to 2.3 log at 100 mg/kg compared to vehicle treated animals. At 25 mg/kg, NI-622 and NI-644 were significantly (P, 0.05) less efficacious than PA-824, however other NI analogs (NI-135, NI-136, NI-182 and NI-297) showed comparable efficacy to PA-824. At 100 mg/kg, AminoEthyl-824, NI-135, NI-136 and NI-302 showed comparable efficacy to PA-824, however NI-622, NI-644, Amino-824 and NI-182 were significantly (P,0.05) less efficacious than PA-824. For PA-824, NI-135 and NI-136 a dose dependent increase in efficacy was observed, on the contrary, no dose-dependent increase in bactericidal activity was observed for NI-622, NI-644 and NI-182. However, none of the selected bicyclic 4-nitroimidazole analogs showed significantly better efficacy than PA-824 at respective 25 and 100 mg/kg doses.

Correlation of PK parameters with efficacy
Further, mouse PK and efficacy data was used to understand the relationship of PK parameters with in vivo efficacy for a series of NI analogs. Both PK and efficacy data at 25 mg/kg were available for only 7 compounds. PK parameters were obtained after a single oral dose (Table 2), while efficacy studies were performed at oral daily doses of 25 and 100 mg/kg for 4 weeks ( Table 4). The relationship between mean log CFU reduction with PK parameters (C max and AUC) was analyzed in both plasma and lungs using the Spearman's rank correlation (Figure 2, Table S1 in File S1). The free plasma concentrations were obtained by correcting for in vitro mouse plasma protein binding. As shown in Figure 2, with the limited set of compounds, the in vivo efficacy correlated well with lung PK parameters than plasma PK parameters. The Spearman's rank correlation coefficient for lung C max and AUC were 0.76 and 0.52 respectively. Across the NI analogs studied, compounds with higher lung concentration (PA-824, NI-297 and NI-182) tended to achieve higher efficacy (Dlog

Correlation of PK-PD indices with efficacy
In vitro activity against Mtb is one of the key determinants of in vivo efficacy, hence the relationship between mean log lung CFU reduction with three primary descriptive PK-PD indices (C max /MIC, AUC/MIC and T .MIC ) was analyzed in both plasma The mean log lung CFU's in five independent in vivo efficacy studies ranged between 6.07 and 6.66 in untreated infected mice. nd = not determined.

D
Mean log lung CFU reduction compared to untreated controls. Each data represents mean value ± SEM from 5 animals.   Table S2 in File S1). As observed above, over all, the in vivo efficacy seems to have strong positive correlation with lung PK parameters than plasma. Among all the PK-PD indices, total lung T .MIC correlated the best with in vivo efficacy (r s = 0.88) than lung C max /MIC (r s = 0.63) and AUC/MIC (r s = 0.63) ( Figure 3C, Table S2 in File S1). For all the compounds analyzed, the total lung T .MIC ranged between 64 and 100% resulting in 0.9-1.56 log lung CFU reduction. In this analysis, NI-644 was found to be an outlier, with T .MIC of 84% resulted in only 0.48 log CFU reduction. Overall, these results suggest that in vivo efficacy of bicyclic 4-nitroimidazole analogs correlates better with the time during which the total lung concentrations are above in vitro potency.

Discussion
Understanding pharmacokinetic-pharmacodynamic (PK-PD) relationships in the early drug discovery process is essential to minimize the attrition rate during the pre-clinical and clinical development phases. In murine models of TB, PK-PD relationships have been established for several standard TB drugs, such as rifampicin [35], isoniazid [36], fluoroquinolones (FQ) [37] and TMC207 [38]. Based on the PK-PD findings with rifampicin, further clinical studies are still in progress to optimize the clinical dose [39][40][41][42]. PA-824, a bicyclic 4-nitroimidazole has demonstrated bactericidal activity in both preclinical and clinical settings [7,8,10]. Extensive medicinal chemistry efforts to improve aqueous solubility, metabolic stability, in vitro potency and in vivo efficacy have independently generated several series of NI analogs [5,[14][15][16][17][18][19][20][21][22][23]. All the bicyclic 4-nitroimidazole analogs analyzed in this study showed cofactor F 420 dependent bio-activation (Table 1) suggesting the mechanism of action of these compounds similar to PA-824 [24,43]. Comprehensive in vivo efficacy studies are generally resource/time intensive and are particularly true for TB. Thus, prioritizing potential lead compounds for in vivo efficacy studies would be useful based on PK parameters. This study is a retrospective analysis of in vivo efficacy with PK for bicyclic 4nitroimidazole analogs to identify the PK parameters and PK-PD indices that correlate with the in vivo potency. The results of this analysis could potentially be exploited to prioritize new analogs for efficacy studies.
Mtb mainly resides in lung granulomatous structures and hence it is important for a drug to be available at the site of the infection for it to be active. The volume of distribution is a primary PK parameter defined by the physico-chemical properties of the compound that indicates the extent of compound distribution in the body. Azithromycin, a macrolide antibiotic, with very high Vss (33 L/kg) is known to have higher lung concentration than serum (AUC lung/serum = 21) and it correlates well with in vivo activity against respiratory pathogens [44,45]. Likewise, moxifloxacin displays a high volume of distribution (Vss = 2 to 5 L/kg) resulting in pronounced penetration into tissues (AUC L/P ratio of 3.3) [46,47] possibly leading to its potent in vivo efficacy against TB [48]. Recently, moxifloxacin has been shown to penetrate and accumulate in granulomatous lesions in TB infected rabbit lungs [49]. TMC207, a diarylquinoline analog, extensively distributes to lungs (AUC L/P ratio of 22) and is efficacious against Mtb [50]. In this study, NI analogs having moderate-to-high volume of distribution (Vss = 1.6 to 4.2 L/kg) and L/P ratio of .2 showed good efficacy in a murine TB model (Dlog CFUs ranging from 1.23 to 1.56) ( Table 2, 3 and 4, Figure 4). Interestingly, NI-622 and NI-644 that showed lower lung to plasma ratio displayed only a marginal efficacy (Dlog CFUs ranging from 0.48 to 0.89). Although, NI-135 and NI-136 showed higher lung to plasma ratio (3.6 to 4.6), their absolute lung concentrations were 2.5 to 7.5 fold lower than PA-824. However, both these compounds displayed 10 times better in vitro potency resulting in comparable in vivo efficacy to PA-824. Overall, the relationship between in vivo efficacy of bicyclic 4-nitroimidazoles displayed positive correlation with Vss (r s = 0.45) (Figure 4). Based on these observations, the Vss and lung distribution could give an initial indication about a compound's potential for in vivo efficacy and thus these two parameters could be used for initial prioritization of compounds during early drug discovery.
A thorough dose fractionation study of PA-824 in a murine model showed that the primary PK-PD driver for in vivo efficacy is the duration during which the free concentration are above MIC (fT .MIC ) in plasma [11]. In this study lung PK parameters have not been measured. Further, fT .MIC in plasma of 22%, 48% and 77% is required for it to show bacteriostatic, 1-log 10 and 1.59 log 10 kill respectively. In general, the PK-PD parameter driving efficacy is conserved within a given class of compounds [51], for example, the efficacy of all FQ analogs is driven by AUC/MIC [37,52,53], while the efficacy of b-lactams correlates with T .MIC [53][54][55]. These studies are done with thorough dose fractionation of single compound with multiple doses and dosing regimen. During lead optimization program prioritization of promising compounds that show good in vivo efficacy is important to reduce the overall turnaround time. In this retrospective analysis with 7 different bicyclic 4-nitroimidazole analogs, we attempted to correlate in vivo efficacy at 25 mg/kg with PK parameters. On the contrary to what has been observed by Ahmad et al., in this study, with seven bicyclic 4-nitroimidazole analogs having varied lung distribution, in vivo efficacy showed weak correlation with free T .MIC in plasma. However, the total T .MIC in lungs showed positive correlation with in vivo efficacy (r s = 0.88) likely due to their preferential distribution into lungs for some analogs. For all the compounds analyzed, the total lung T .MIC ranged between 64-100% resulting in 0.9-1.56 log lung CFU reduction, hence efficacy studies at lower doses (resulting in T .MIC less than 65%) might be necessary to see a better correlation. Overall, in this study a diverse set of bicyclic 4-nitroimidazoles with Vss ranging from 0.7 L/kg to 4.2 L/kg, lung to plasma ratio ranging from 0.5 to 4.6 showed positive correlation with lung T .MIC than with any other parameters.
The results presented in this study must be interpreted with a couple of limitations in mind. First, single-dose PK parameters determined in healthy mice were assumed to be similar to multiple-dose PK parameters in infected animals and were correlated with efficacy data. This assumption is supported by published preclinical data that has shown the absence of plasma accumulation of PA-824 in mice dosed for 2 months [56]. Further, in clinical studies with PA-824, the PK parameters from a single dose phase I study were similar to a multiple dose phase II study in patients [9,10,12,13]. Another limitation of this study is that total concentrations in lungs rather than the free lung concentrations were used for the PK-PD analysis. It is well accepted that for a given compound unbound drug concentrations in plasma are equivalent to unbound tissue concentrations when active transport is not involved in the drug distribution [57,58]. Further, it is the unbound concentration of a compound at its target site driving the pharmacological effect [58][59][60][61][62][63]. Nevertheless, whole-tissue concentrations can be of some value in early drug discovery providing a first assessment of partition into the lungs [61]. Techniques like microdialysis in lungs can be applied to assess unbound tissue concentration [64,65]. In TB patients, Mtb mainly resides in diverse and heterogeneous lesions in lungs. In general, interpretation of PD activity of anti-TB compounds is complicated by differential lung pathophysiology. PK in intrapulmonary compartments like the epithelial lining fluid and alveolar macrophages have also been studied in humans for standard TB drugs like rifampicin [66], isoniazid [67], ethambutol [68], pyrazinamide [69], rifapentine [70], moxifloxacin [71], ofloxacin [72] and linezolid [73]. The concentration in these sites could be the key factor governing the efficacy of anti-TB drugs. However, measurement of compound concentration in lungs by microdialysis, epithelial lining fluid and alveolar macrophages have limitations in sampling, methodology and interpretation of results [61,74]; and such studies have not been explored in preclinical settings for TB. The total lung concentration may not be equal to the concentration in Mtb lesions, thus warranting lesion PK analysis to improve the predictive power for efficacy. Recently PK in lung lesions of mycobacterium-infected rabbits has been investigated for isoniazid, rifampicin, pyrazinamide and moxifloxacin [49,75]. Although lesion PK can offer better insights in understanding PK-PD relationships, it is not easily applicable to early drug discovery especially with mouse efficacy model as it doesn't display spectrum of lesions observed in TB patients or in higher animal models. In addition, similar studies with bicyclic 4nitroimidazoles may be challenging as they undergo enzymatic transformation in Mtb to multiple stable and unstable metabolites [43].
Our findings show that the efficacy of all bicyclic 4nitroimidazole analogs is most likely driven by PK parameters in lungs. A simple efficacy surrogate would be useful during the lead optimization to prioritize candidates for lengthy efficacy studies. For this class, efficacy correlated better with concentration in lungs rather than in plasma, consistent with Vss and differential lung: plasma distributions. The results of this analysis potentially be exploited to prioritize new analogs for efficacy studies based on in vitro potency, volume of distribution and lung concentration.

Supporting Information
File S1 Figure S1: Plasma concentration time profile for representative bicyclic 4-nitroimidazole analogs following a single 25 mg/kg oral dose in mice. Table S1: Correlation of PK parameters with in vivo efficacy in mice for bicyclic 4-nitroimidazole analogs. Table S2: Correlation of PK-PD indices with in vivo efficacy in mice for bicyclic 4-nitroimidazole analogs. (DOCX)